Dale Tovar creates new sparse array format for SciPy

There are countless numbers of applications which involve large amounts of sparse data in all scientific and commercial domains. To that end, our own Dale Tovar has created and implemented the GXCS format, which is a novel storage scheme for ultra-large sparse arrays and tensors. One of the advantages of the new format is that it takes up much less storage space in many of the scenarios encountered in practice. For example, unlike the commonly used COO format where storage size increases strongly as both a function of the number of axes and the number of nonzero values, GXCS increases in storage size primarily as a function of the number of nonzero values. The software includes conversions to and from different formats. This will help greatly speed up analyses for anyone who uses it. 

In even bigger news, GXCS is written in the Python and has been incorporated in to pydata/sparse library which is replacing the defacto sparse matrix library in SciPy in a couple of years!

Amazing work and huge congrats to Dale!

Rob Chavez

Assistant Professor of Psychology at the University of Oregon and Director of the Computational Social Neuroscience Lab.

Leave a Reply

Your email address will not be published. Required fields are marked *